China’s DeepSeek developing its own AI chip to cut reliance on Nvidia, Huawei: Report

China’s DeepSeek is developing its own AI chip aimed at AI inference, reports Reuters citing three people familiar with the matter. The move comes as the Chinese startup looks to reduce its reliance on chips from Nvidia and Huawei.

As per the report, the effort is still at an early stage, with the company working with external partners and expanding its chip design team. If successful, the move would mark a major shift for DeepSeek, which has become one of China's leading AI companies after its AI models gained global attention.

DeepSeek focuses on inference chip

According to Reuters, DeepSeek's chip is being designed for AI inference, the stage where a trained AI model generates answers for users, rather than for training new AI models.

The report said the Hangzhou-based company began the effort about a year ago. It has been speaking with chip-design, foundry and memory companies while quietly hiring more chip-design engineers without posting public job listings, according to two of the sources.

Move follows global AI trend

Reuters reported that DeepSeek's chip project would bring it in line with other AI companies seeking greater control over the hardware used to power their models.

Last month, OpenAI introduced its first custom inference chip, Jalapeno, developed with Broadcom. Reuters also reported in April that Anthropic had been considering building its own AI chips.The report said the project also reflects the impact of U.S. export controls, which prevent Chinese companies from buying Nvidia's most advanced AI chips.DeepSeek founder Liang Wenfeng said in a rare 2024 interview with a Chinese media outlet that "chip export controls were a challenge for the company."DeepSeek has used both Nvidia and Huawei chips in the past. The company has said the foundation model behind its R1 reasoning model was trained on Nvidia's H800 chip, which was designed for China before Washington banned its exports in late 2023.According to the report, DeepSeek has since relied more heavily on Huawei. In April, it released its V4 model optimized for Huawei's Ascend chips, and Huawei said its processors were used for part of the training of the V4-Flash model.